mock-llm-provider. Scripted OpenAI Responses mock backend. Replay streaming, reasoning, and tool calls from YAML scenarios to reproduce client bugs deterministically.